The Japanese government has been attempting to reform the national research system for the past 20 years. This paper describes the structural changes of the system and its performance based on bibliometric analyses and discusses the effects of S&T policy. The investigation indicates that although Japan gradually increased its production of highly cited publications, its share of low-cited publications is much higher than the former. Detailed analyses reveal that the top eight universities account for half of the highly cited publications in the university sector, while other hundreds of universities have massively increased their low-cited publications since 1990. The development of financial and human resources for research in the 1990s enabled new actors to be involved in scientific research, but the resources were concentrated to a small number of universities, reinforcing the collaboration between these universities and others.